Genetically Depauperate in the Continent but Rich in Oceanic Islands: Cistus monspeliensis (Cistaceae) in the Canary Islands Mario Ferna´ndez-Mazuecos*, Pablo Vargas Real Jardı´n Bota´nico, CSIC, Madrid, Spain
Abstract Background: Population genetic theory holds that oceanic island populations are expected to have lower levels of genetic variation than their mainland counterparts, due to founder effect after island colonization from the continent. Cistus monspeliensis (Cistaceae) is distributed in both the Canary Islands and the Mediterranean region. Numerous phylogenetic results obtained in the last years allow performing further phylogeographic analyses in Cistus. Methodology/Principal Findings: We analyzed sequences from multiple plastid DNA regions in 47 populations of Cistus monspeliensis from the Canary Islands (21 populations) and the Mediterranean basin (26 populations). The time-calibrated phylogeny and phylogeographic analyses yielded the following results: (1) a single, ancestral haplotype is distributed across the Mediterranean, whereas 10 haplotypes in the Canary Islands; (2) four haplotype lineages are present in the Canarian Islands; (3) multiple colonization events across the archipelago are inferred; (4) the earliest split of intraspecific lineages occurred in the Early to Middle Pleistocene (,930,000 years BP). Conclusions/Significance: The contrasting pattern of cpDNA variation is best explained by genetic bottlenecks in the Mediterranean during Quaternary glaciations, while the Canarian archipelago acted as a refugium of high levels of genetic diversity. Active colonization across the Canarian islands is supported not only by the distribution of C. monspeliensis in five of the seven islands, but also by our phylogeographic reconstruction in which unrelated haplotypes are present on the same island. Widespread distribution of thermophilous habitats on every island, as those found throughout the Mediterranean, has likely been responsible for the successful colonization of C. monspeliensis, despite the absence of a longdistance dispersal mechanism. This is the first example of a plant species with higher genetic variation among oceanic island populations than among those of the continent. Citation: Ferna´ndez-Mazuecos M, Vargas P (2011) Genetically Depauperate in the Continent but Rich in Oceanic Islands: Cistus monspeliensis (Cistaceae) in the Canary Islands. PLoS ONE 6(2): e17172. doi:10.1371/journal.pone.0017172 Editor: Wayne Delport, Prognosys Biosciences, United States of America Received September 23, 2010; Accepted January 24, 2011; Published February 14, 2011 Copyright: ß 2011 Ferna´ndez-Mazuecos, Vargas. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: The research on continental plants has been supported by the Spanish Direccio´n General de Investigacio´n Cientı´fica y Te´cnica (DGICYT) through project CGL2005-06017-C02 and a PhD scholarship to M.F.-M. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * E-mail:
[email protected]
been paid to populations of species distributed both on continents and oceanic islands. Nonetheless, populations of the same species distributed in insular and mainland areas can provide key insights into microevolutionary processes underlying recent colonization and early stages of differentiation. As a general pattern, lower levels of genetic variation are expected in island populations as compared to mainland populations due to founder effects and restricted gene flow [9]. This depauperation may bring about an increased propensity for extinction and compromised evolutionary potential in island populations. Early studies of genetic variation in mainland and island populations included only one study of oceanic islands [9,10]. More recent examples of islandmainland comparisons have been reported, most of which also focused on continental islands [11,12,13, but see 14,15]. These examples generally agree with the expectation of higher genetic variation in mainland populations. The finding of higher variation in continental islands has been attributed to multiple continent-to-island introductions, genetic bottlenecks in the
Introduction Islands constitute a focus of research interest in plant evolutionary biology, given their limited area and varying degrees of isolation from nearby continents. Continental islands are located on continental shelves, and were isolated from the continent by means of rising sea level or/and by tectonic processes. These islands may have been recurrently connected to the continent by land bridges due to fluctuating sea levels. On the contrary, oceanic islands arise from the ocean floor, are usually of volcanic origin and have virtually no terrestrial life in origin [1,2]. Because they furnish clear-cut spatial and temporal limits, oceanic islands are considered to be living laboratories for evolution. That is why oceanic islands provide ideal systems to investigate historical colonization and evolutionary patterns in plants [3]. Speciation processes giving rise to endemic species and lineages on oceanic islands have been widely discussed, leading to alternative models of evolution [2,4,5,6,7,8]. Less attention has PLoS ONE | www.plosone.org
1
February 2011 | Volume 6 | Issue 2 | e17172
High Plastid Variation in a Canarian Cistus
Islands (53 individuals; Table 1), plus 26 Mediterranean populations (26 individuals) previously analyzed [30] (Table S1). All new individuals were collected in the field and dried in silica gel. First, a pilot study was performed to find the most variable sequences among 17 plastid DNA regions previously used in phylogenetic and phylogeographic analyses [see 30 for details]. Procedures used for amplification and sequencing of DNA regions followed Ferna´ndez-Mazuecos and Vargas [30]. After identifying two DNA regions with high sequence variation (trnStrnG and psbK-trnS), we extended the sequencing to every population (1–3 individuals per population in the Canary Islands). The same DNA regions were also sequenced from the other eleven species of the white-flowered lineage [19,32] and two purpleflowered species as the outgroup (Table S1), to reconstruct phylogenetic relationships of the cpDNA haplotypes (see below). A recently described species (C. grancanariae [29]) related to C. monspeliensis needs further taxonomic validation. Nevertheless, a preliminary study rendered no sequence variation between the two taxa for the two cpDNA regions (Ferna´ndez-Mazuecos & Vargas, unpublished). All new sequences have been deposited in GenBank (see Table S1 for accession numbers).
continent, or island-to-continent colonization [13,16,17]. To our knowledge, higher genetic variation has not been reported for oceanic islands, where particularly strong genetic bottlenecks are expected due to isolation and the prevalence of single introduction events [18]. The genus Cistus L. (Cistaceae) comprises 21 species of primarily Mediterranean distribution. The highest species diversity is found in the western Mediterranean, which has 14 species. Seven species are present in the Canary Islands (Macaronesian region). Five of them are endemic and form a sublineage within the purpleflowered clade [19,20], while the white-flowered C. monspeliensis and C. ladanifer are also widely distributed Mediterranean elements. Although a close relationship between the Mediterranean and Macaronesian floras is well established [8,21,22,23], there are few reports of phylogeographic and population genetic analysis of species that are present in both floristic regions [but see 14,24,25], such as these two Cistus species. This might be due to the risk of including plant populations introduced since the first human (guanches) colonization, coupled with more interest in the endemic element [4,6,26]. In particular, it is not clear whether the occurrence of C. monspeliensis in the Canary Islands is the result of natural or human-mediated introduction [19]. Some white-flowered species of Cistus with a local presence in Macaronesia are considered introduced species, such as C. ladanifer in the Canary Islands [27] and C. psilosepalus and C. salviifolius in Madeira [28]. In contrast, it has been suggested that C. monspeliensis is a native species in the Canary Islands [23], although no morphological differentiation has been reported for most of the populations that supports this hypothesis (but see C. grancanariae [29]). A recent phylogeographic analysis found no nucleotide variation in 26 Mediterranean populations of C. monspeliensis after screening 17 cpDNA regions, which indicated that there was a rapid dispersal across the Mediterranean after species formation in the Pleistocene [30]. The question remains as to whether Canarian populations share the same Mediterranean genotype or show some degree of genetic exclusiveness, the latter of which would support the native status hypothesis. The plastid genome is structurally stable, haploid, nonrecombinant and maternally inherited in Cistus [31]. Accordingly, it has been used to infer phylogeographic and seed colonization patterns for this genus [20,30,31]. In the present study, we analyzed cpDNA haplotypes of insular and continental populations of C. monspeliensis. Our goal was three-fold: (1) to compare levels of genetic variation; (2) to determine native versus humanintroduced status for the Canarian populations; and (3) to reconstruct the colonization history in the Canary Islands.
Haplotype data analyses All trnS-trnG and psbK-trnS sequences were assembled in Geneious Pro 4.8.3 [33] and aligned using ClustalW 2.0.12 [34]. Further adjustments were made by visual inspection. Species most closely related to C. monspeliensis according to previous results were included as the outgroup. Genealogical relationships among haplotypes based on nucleotide substitutions were inferred using the statistical parsimony algorithm [35], as implemented in TCS 1.21 [36]. The maximum number of differences resulting from single substitutions among haplotypes was calculated with 95% confidence limits, treating gaps as missing data. Phylogenetic relationships were also assessed using maximum parsimony (MP) and Bayesian inference (BI). Analyses were conducted combining sequences of the two DNA regions representing the haplotypes of C. monspeliensis plus the other eleven white-flowered and two purple-flowered Cistus species. MP analyses were performed using PAUP* 4.0b10 [37], with the following parameters for the heuristic search: 1000 random addition replicates holding 100 trees per replicate, tree-bisectionreconnection (TBR) branch swapping and the options Multrees and Steepest Descent in effect. Robustness of clades was estimated using 1,000,000 bootstrap replicates (fast stepwise-addition [38]). BI was implemented in MrBayes v3.1.2 using two identical searches with 10 million generations each (chain temperature = 0.2; sample frequency = 100). The simplest model of sequence evolution that best fits the sequence data (GTR+G) was determined under the Akaike Information Criterion (AIC) in jModeltest 0.1.1 [39,40]. Probabilities converged on the same stable value after c. 20,000 generations in both runs. A 50% majority rule consensus tree was calculated to obtain the Bayesian estimate of phylogeny [see 19 for details].
Methods Study species Cistus monspeliensis L. is a lowland shrub displaying a rather continuous distribution in the Mediterranean basin. It usually occurs on poor soils of the thermomediterranean vegetation belt (600–800 m) on both calcareous and acidic substrates. In the Canary Islands, it occurs in the understory and successional scrub of thermophilous forests, laurel forests and Pinus canariensis woodlands of five islands: Tenerife, Gran Canaria, La Palma, El Hierro and La Gomera [23]. It has also been reported in Madeira, but it is probably no longer present in this archipelago [28].
Genetic diversity For each island, we calculated the number of haplotypes based on nucleotide substitutions (h), number of private haplotypes (ph) and haplotypic diversity (H) [41]. The same parameters were computed for the Mediterranean basin and the Canary Islands as a whole. Using previously published cpDNA data [20,30,31], h and H were calculated for other monophyletic Cistus species and lineages recently differentiated in the Mediterranean and the Canary Islands. Given that the trnS-trnG spacer has been employed in all published analyses of Cistus, h and H were also calculated for
Sample strategy and DNA sequencing A total of 47 populations of Cistus monspeliensis were sampled to cover its distributional area: 21 populations from the Canary PLoS ONE | www.plosone.org
2
February 2011 | Volume 6 | Issue 2 | e17172
High Plastid Variation in a Canarian Cistus
Table 1. Canarian Cistus monspeliensis populations used for plastid sequencing of the trnS-trnG and psbK-trnS regions, and haplotypes found in sequenced individuals.
Population number
Locality (number of individuals)
Voucher
Haplotype codes
1
El Hierro, La Pen˜a (3)
C. Garcı´a-Verdugo 39CG05 (MA)
B/B/B
2
El Hierro, Valverde (3)
C. Garcı´a-Verdugo 41CG05 (MA)
F/J/J
3
La Palma, Barranco Garome (3)
C. Garcı´a-Verdugo 34CG05 (MA)
C/C/D
4
La Palma, La Tosca (3)
B. Guzma´n 142BGA04 (MA)
C/C/C
5
La Palma, Los Llanos – Santa Cruz (3)
V. Valca´rcel 63VV04 (MA)
C/C/D
6
La Palma, Villa de Mazo (1)
P. Vargas 254PV02 (MA)
C
7
La Palma, Barranco Seco (3)
C. Garcı´a-Verdugo 26CG05 (MA)
D/D/D
8
La Gomera, Arure (3)
C. Garcı´a-Verdugo 19CG05 (MA)
H/K/K
9
La Gomera, Alto de Garajonay (3)
A. Herrero AH2443 (MA)
J/J/J
10
La Gomera, Jaraga´n (2)
C. Garcı´a-Verdugo 10CG05 (MA)
J/J
11
Tenerife, Barranco de las A´nimas (1)
B. Guzma´n 12BGA05 (MA)
B
12
Tenerife, Villa de Arico (3)
P. Vargas 42PV05 (MA)
E/E/H
13
Tenerife, road to Teide (1)
B. Guzma´n 14BGA05 (MA)
G
14
Tenerife, Aguamansa (1)
C. Garcı´a-Verdugo 8CG05 (MA)
J
15
Tenerife, Gu¨´ımar (3)
C. Garcı´a-Verdugo 54PV05 (MA)
J/J/J
16
Tenerife, Igu¨este de San Andre´s (3)
P. Vargas 56PV05 (MA)
B/B/B
17
Gran Canaria, Embalse del Mulato (2)
B. Guzma´n 1BGA05 (MA)
G/G
18
Gran Canaria, Artenara (3)
P. Vargas 61PV05 (MA)
B/G/J
19
Gran Canaria, road GC-605 (3)
B. Guzma´n 3BGA05 (MA)
I/I/J
20
Gran Canaria, San Bartolome´ de Tirajana (3)
B. Guzma´n 6BGA05 (MA)
G/G/G
21
Gran Canaria, Roque Nublo (3)
P. Vargas 169PV08 (MA)
G/G/J
doi:10.1371/journal.pone.0017172.t001
this region separate. In all cases, calculations were performed in DnaSP v5 [42]. In this software, sites with alignment gaps (or missing data) are excluded from calculations. Therefore, we eliminated certain samples (one individual of C. creticus and one individual of C. symphytifolius [20]) with indels or missing data in variable sites in order to properly infer substitution-based diversity parameters.
[46,47,48]. Unlike other modern approaches to phylogeographic inference [49], these models do not infer the demographic history of populations, but they are able to reconstruct historical movements of or between populations [46]. Therefore, this methodology is appropriate for our inference of colonization history across an oceanic archipelago. We employed C. populifolius as outgroup based on previous phylogenies [30,32], and the HKY substitution model was chosen following jModeltest result. We implemented a relaxed molecular clock, with an uncorrelated lognormal distribution for the substitution rate variation, and a coalescent model with constant size was assumed as tree prior. The root height was modelled as a normal distribution with mean = 1.13 Ma, based on the divergence time between C. monspeliensis and C. populifolius previously estimated using fossil calibrations [30]. The uncertainty on this calibration point (95% highest posterior density interval 0.24–2.41 Ma [30]) suggests a standard deviation = 0.66. However, a stronger prior is desirable for the root height in order to estimate evolutionary rates, and thus, with values above 0.20, the MCMC failed to converge to a coherent result, with all node ages approaching zero. Therefore, we chose 0.20 as an appropriate value of standard deviation, but we acknowledge that the uncertainty on divergence times may be higher than shown by our results. In any case, our ultimate goal was to find a reliable upper bound rather than a precise dating. The colonization history was reconstructed using a Bayesian phylogeographic framework [47]. We defined six areas (the five islands plus the Mediterranean region), and they were mapped with a discrete phylogeographic analysis using a standard continuous-time Markov chain (CTMC). As suggested by [47], we also implemented a Bayesian stochastic search variable
Among-island genetic differentiation The nearest-neighbour statistic (Snn) was calculated to assess genetic differentiation in C. monspeliensis due to isolation in different islands. This statistic is a measure of how often the ‘‘nearest neighbours’’ (in sequence space) of sequences are from the same locality in the geographic space [43]. Snn is expected to approach one when two partitions (areas, localities) of a dataset form highly differentiated populations, and one-half when they are part of a single panmictic population. The dataset was divided into five partitions corresponding to the five islands with presence of the species. Snn was calculated using DnaSP v5, and permutation tests with 1000 replicates were performed to evaluate significance of the obtained values.
Bayesian dating and phylogeographic reconstruction In order to infer divergence times among C. monspeliensis lineages and reconstruct its colonization history across the Canarian archipelago, the dataset was analyzed using a relaxed Bayesian approach as implemented in BEAST v.1.6.1 [44,45]. We employed the spatial diffusion methodogy, a recently developed approach to phylogeography aimed to identify the ancestral geographical history of a sample of molecular sequences PLoS ONE | www.plosone.org
3
February 2011 | Volume 6 | Issue 2 | e17172
High Plastid Variation in a Canarian Cistus
clades are connected to haplotype B. The first one is a lineage exclusive to La Palma (C–D). The second and third lineages are tip haplotypes connected to B and only found in Tenerife (E) and El Hierro (F). The fourth lineage is highly differentiated (five haplotypes: G–H–I–J–K) and widely distributed in all islands except for La Palma. Interestingly, unrelated haplotypes are found in the young island of El Hierro (B–F, J) and also in La Gomera (H, J–K).
selection (BSSVS) procedure to identify parsimonious descriptions of the diffusion (colonization) process. We employed all the standard parameters suggested in the authors’ web site (http:// beast.bio.ed.ac.uk/Tutorials). Two MCMC analyses were run for 10 million generations, sampling every 1000th generation. Analysis with Tracer 1.4 [50] confirmed convergence among chains and adequate sample size. Both chains were combined using LogCombiner 1.4.8 after discarding the first 10% of sampled generations as burn-in, and trees were summarized in a maximum clade credibility (MCC) tree obtained in TreeAnotator 1.6.1 and visualized in FigTree 1.3.1. Finally, a Bayes factor (BF) test was performed to identify rates (colonization routes) that are frequenly invoked to explain the diffusion process. Rates yielding a BF.3 were considered as well supported, and were converted into a KML file suitable for visualization in Google Earth [47].
Phylogenetic analysis Combination of trnS-trnG and psbK-trnS sequences of 14 Cistus species (including those of the eleven haplotypes of C. monspeliensis) resulted in an aligned length of 1037 bp. Forty-two of the 73 variable sites from the matrix were phylogenetically informative. MP analysis generated 1322 trees of 89 steps with a consistency index (CI) of 0.88 and a retention index (RI) of 0.93. The strict consensus tree (Fig. 1D) recognizes C. monspeliensis populations as monophyletic, with a 99% bootstrap support (BS). The BI phylogeny depicted a congruent topology, with a posterior probability (PP) of 1.00 for the clade of haplotype sequences of the study species. Phylogenetic relationships among haplotypes are congruent with those retrieved in the network analysis. The Mediterranean haplotype (A) is recovered as sister to the Canarian clade, which is supported as monophyletic (60% BS; 0.92 PP).
Results Haplotype analysis Among the 17 cpDNA regions tested, two spacers (trnS-trnG and psbK-trnS) showed the highest levels of variation. Length of the aligned sequences without the outgroup was 617 bp for trnS-trnG and 367 bp for psbK-trnS. The combined analysis of the two regions yielded eleven substitution-based haplotypes of C. monspeliensis (Table 2) distributed in the Canary Islands (ten haplotypes, Fig. 1B, Table 1) and the Mediterranean region (one haplotype, Fig. 1A). Haplotype A was found in the 26 Mediterranean populations, as previously reported [30]. Within the Canary Islands, six of the ten haplotypes were exclusive to single islands: C and D to La Palma, E to Tenerife, F to El Hierro, I to Gran Canaria and K to La Gomera. The remaining four haplotypes were shared by two or more islands: B by Tenerife, Gran Canaria and El Hierro, G by Tenerife and Gran Canaria, H by Tenerife and La Gomera, and J by Tenerife, Gran Canaria, La Gomera and El Hierro. In the TCS analysis (Fig. 1C) all Canarian haplotypes of C. monspeliensis formed a single network with no loops, which was connected to the Mediterranean haplotype (A) through haplotype B. Two mutational steps separated haplotypes A and B. Four
Genetic diversity Tenerife was found to harbour the highest diversity measured in haplotype number (h = 5). This island also had the highest haplotypic diversity (H = 0.803), followed by El Hierro, Gran Canaria, La Gomera and La Palma (Table 3). Despite showing the lowest diversity, La Palma contained two private haplotypes, while the remaining islands contained one private haplotype each. These results and the high diversity found in the Canary Islands as a whole (h = 10; H = 0.857) contrast with the lack of diversity in the Mediterranean basin (h = 1; H = 0.000). When comparing haplotype number and haplotypic diversity in monophyletic Cistus lineages (Table 4), C. monspeliensis populations from the Canary Island yielded the highest values in both estimators. The high cpDNA haplotypic diversity of Canarian C.
Table 2. Variable sites of substitution-based haplotypes found in 47 C. monspeliensis populations from the Mediterranean (A) and the Canary Islands (B–K), based on trnS-trnG (617 bp) and psbK-trnS (367 bp) sequences.
DNA region
trnS-trnG sequence position
psbK-trnS sequence position
Haplotype
Number of individuals
A
26
T
G
G
A
T
A
A
T
A
C
T
A
B
8
T
A
T
A
T
A
A
T
A
C
T
A
C
8
T
A
T
A
T
A
A
T
A
C
A
A
D
5
T
A
T
A
T
A
A
T
T
C
A
A
E
2
T
A
T
A
T
A
A
T
A
A
T
A
F
1
T
A
T
A
T
A
C
T
A
C
T
A
G
9
T
A
T
A
T
A
A
G
A
C
T
A
H
2
T
A
T
A
T
C
A
G
A
C
T
A
I
2
T
A
T
C
G
A
A
G
A
C
T
A
J
14
T
A
T
C
T
A
A
G
A
C
T
G
K
2
G
A
T
C
T
A
A
G
A
C
T
G
3
26
357
358
396
412
458
6
14
93
124
365
doi:10.1371/journal.pone.0017172.t002
PLoS ONE | www.plosone.org
4
February 2011 | Volume 6 | Issue 2 | e17172
High Plastid Variation in a Canarian Cistus
PLoS ONE | www.plosone.org
5
February 2011 | Volume 6 | Issue 2 | e17172
High Plastid Variation in a Canarian Cistus
Figure 1. Phylogeographic analysis of Cistus monspeliensis based on cpDNA sequences. Sampled populations of Cistus monspeliensis in the Mediterranean (A) and the Canary Islands (B) indicating geographical location of the 11 cpDNA haplotypes (colours) inferred from sequences of the trnS-trnG and psbK-trnS regions. Chart sizes in B are proportional to the number of sequenced individuals. Canarian populations are numbered as in Table1. Maps: SRTM Shaded Relief, ESRI. (C) Statistical parsimony network of C. monspeliensis haplotypes (indicated by letters) and six closely related species. Lines represent single nucleotide substitutions; dots indicate absent haplotypes (extinct or not found). Colours are as depicted for A and B. (D) Strict consensus tree of the 1322 shortest trees of 89 steps (CI = 0.88; RI = 0.93) from the combined analysis of trnS-trnG and psbK-trnS sequences. Numbers above branches are bootstrap values; numbers below branches are Bayesian posterior probabilities. doi:10.1371/journal.pone.0017172.g001
monspeliensis (H = 0.857) is only matched by the Canarian purpleflowered lineage of five species (H = 0.832). When analysing the trnS-trnG spacer alone, the highest number of haplotypes (h = 6) was found in Canarian populations of C. monspeliensis and the Western Mediterranean C. ladanifer (Table 4). We found similarly high values of H for trnS-trnG sequences in the Canarian C. monspeliensis (H = 0.572), Canarian purple-flowered Cistus (H = 0.591) and Mediterranean C. laurifolius (H = 0.589). Mediterranean C. monspeliensis populations constitute the only analyzed Cistus lineage without cpDNA variation of substitutionbased haplotypes found to date. A very low variation is also found in C. salviifolius (h = 2; H = 0.208).
ceived the highest probability (0.31). Two islands were similarly supported as the most likely ancestral range of all Canarian samples: Gran Canaria (0.32) and Tenerife (0.29). The ancestor of the lineage formed by haplotypes G-K probably inhabited Gran Canaria (0.61) no longer than 350,000 years ago. Despite uncertainty on topology and direction of colonization events, the MCC tree supports a single colonization of La Palma, two colonizations of El Hierro, two colonizations of La Gomera and several exchanges between Gran Canaria and Tenerife. Three migration routes were supported by the BF test (Fig. 2): Gran Canaria-Tenerife (BF = 23.20), Tenerife-La Gomera (BF = 9.68) and Tenerife-El Hierro (BF = 3.74).
Genetic differentiation
Discussion
Values of Snn are shown in Table 5. Significant genetic differentiation was retrieved in seven island comparisons: La Palma-Tenerife, La Palma-Gran Canaria, La Palma-El Hierro, La Palma-La Gomera, Tenerife-Gran Canaria, Gran Canaria-El Hierro and Gran Canaria-La Gomera. The highest significant values (Snn
